Speaker
Description
Adaptive mesh refinement (AMR) offers a practical route toward digital twins in computational fluid dynamics, automatically tailoring mesh resolution for each full-order simulation to maintain accuracy across large ensembles with varying boundary conditions and geometries. This work investigates the coupling of Scale-Adaptive Unsteady Reynolds-Averaged Navier-Stokes (SA-URANS) modelling with AMR to improve the trade-off between computational cost and predictive accuracy in turbulent flow simulations, and evaluates the proposed methodology on canonical urban microclimate configurations. Turbulence-driven refinement strategies are implemented in OpenFOAM v13 through extensions of the dynamic-mesh infrastructure, with refinement indicator fields computed on the fly during simulations. In addition to a conventional Q-criterion-based strategy, novel criteria based on the von Kármán length scale are introduced to concentrate resolution where scale-resolving behaviour is expected. The results indicate that turbulence-driven AMR enables simulations to begin from a coarse, far-from-optimal computational grid and progressively increase effective resolution under fixed resource constraints. This provides a flexible, automation-compatible framework that supports the use of SA-URANS with AMR as a scalable building block for digital twins.